Does the Medicare Principal Inpatient Diagnostic Cost Group Model Adequately Adjust for Selection Bias?

Abstract

This study examines bias in the Medicare Principal Inpatient Diagnostic Cost Group (PIP-DCG) model due to unobserved selection using HMO and FFS hospital use data. It found that unobserved selection is systematically different in the FFS and HMO populations, with HMO enrollees healthier and FFS beneficiaries sicker in ways not captured by the PIP-DCG model. As a result, the FFS-based model overestimates HMO enrollees' health care resource use compared to their use if they had been served in FFS. This research should be of interest to researchers and policymakers who are interested in risk adjustment methodologies and are concerned with Medicare overpayments to Medicare+Choice health plans. Previous researchers have not been able to assess the PIP-DCG model using actual utilization data of HMO enrollees. The unique data set in this study comes from a RAND research project that linked California hospital discharge data to Medicare administrative data. The dissertation was completed in partial fulfillment of the requirements of the RAND Graduate School for the degree of Doctor of Philosophy in Policy Analysis.

Open PDF

Document Details

Document Type
Technical Report
Publication Date
Jan 01, 2002
Accession Number
ADA411881

Entities

People

  • Hongjun Kun

Organizations

  • RAND Corporation

Tags

Communities of Interest

  • Biomedical

DTIC Thesaurus Topics

  • Abstracts
  • Availability
  • California
  • Classification
  • Contracts
  • Copyrights
  • Data Sets
  • Health Care
  • Health Services
  • Hospitals
  • Intellectual Property
  • Law
  • Medicare
  • Monitoring
  • Philosophy
  • Theses

Fields of Study

  • Political science

Readers

  • Medical or Health Care Field.
  • Psychometric Testing or Psychological Assessment.